AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
RL in Chat GPT
In our very preliminary work, you don't even have to do a full arm. We're applying it to two domains. One is how can you watch humans perform a task for us? It's a packing task. How can robots watch what a human is doing and extract that information about how they repositioned an object,. And then this other aspect too, which is very interesting is, well, if you want to efficiently learn a demonstration, perhaps with some human input, how do you ask the human an efficient question? So there is strict RL in terms of just adapting. Many times your RL is explore and exploit. But sometimes exploring can be dangerous. If I have a moving robot